Agent engineering

AI Agent Development

Build tool-using agents with MCP, function calling, retrieval, memory and reliable multi-step execution patterns.

Move from a prompt demo to an observable agent workflow.

Learn in the order developers build.

Start with the decision, move into implementation and finish with production checks.

Structured learning path

Define the task

Set boundaries, completion criteria and human checkpoints.

Connect tools

Use tool calling or MCP with strict schemas and permissions.

Add context

Combine retrieval, memory and stable cached instructions.

Evaluate runs

Measure completion, corrections, latency and cost per task.

Build topical authority, not article islands.

Core guides

5 guided starting points

Foundation01

AI Agent API Fundamentals

The request loop behind planning, tools and completion.

Open guide

Protocol02

MCP Integration Guide

Connect agents to tools and data with Model Context Protocol.

Open guide

Tools03

Tool Calling Patterns

Schemas, validation, retries and safe execution.

Open guide

Context04

RAG for Agents

Retrieve the minimum useful context for each step.

Open guide

Memory05

Agent Memory Design

Separate working memory, user memory and durable state.

Open guide

Build a repository agent

Give an agent a bounded goal, repository tools and explicit verification steps before it reports completion.

Read-only discovery first

Scoped write permissions

Tool-result validation

Build and test evidence

Turn the guide into a request.

Try a model in the Playground, then use the Quickstart to move the same pattern into Python or Node.js.

Try without coding Open Quickstart

Explore every growth topic.

AI API Guides

Go from model shortlist to a working production request.

LLM Infrastructure

Build an AI request layer that survives provider and model changes.

AI Agent

Move from a prompt demo to an observable agent workflow.

Cost Optimization

Lower cost without hiding quality or reliability trade-offs.

Model Comparison

Select a model based on the job, not the leaderboard headline.

Tutorials

Copy a working pattern, understand it and adapt it to your application.

Frequently asked questions

AEO-ready answers

What makes an AI agent different from a chatbot?

An agent can choose and execute tools across multiple steps while tracking state toward a defined outcome.

When should an agent use memory?

Use memory only when information must persist across steps or sessions; keep transient reasoning and durable user facts separate.